Re: Software RAID-5 attempt to access beyond end of device...

On Mon, 2004-08-30 at 08:27, Neil Brown wrote:

On Friday August 27, su@rotamente.com wrote:

quoted

Hello,

I have a problem with an Software RAID-5 at the top of 4 IDE-Drives
connected via 2 Promise UltraTX2 (20268) IDE controller.

This is not a raid problem.  It is a problem with your filesystem or
your hardware.
I think you have reported this for both reiserfs and ext3, in which
can it is unlikely to be the filesystem.  I guess it could be in the
generic block layer...

quoted

FS is EXT3.
Kernel 2.6.8.1	(problem occured with 2.6.7, update didn't helped)
CPU: Pentium II 400
Mainboard: Asus P2B-S

After 29% usage, I've got an Kernel oops and system stalls:
-------- snip start -------------
Aug 27 16:30:18 gharb kernel: attempt to access beyond end of device
Aug 27 16:30:18 gharb kernel: md0: rw=1, want=6917529027856096120,
limit=241215360

6917529027856096120 is 600000000CD0DB78 in hex.
0xCD0DB78 is 215014264 which is not beyond the end of the device.
It looks like  2 extra bits have been set at the top of the address.
This could be due to bad memory (though that is more commonly a
single-bit error, not a double bit error), or it could be the
filesystem send a bad request, possibly due to bad indexing
information that has previously been stored on disk.

I recommend running memtest for a while and confirming that it isn't a
hardware error, and then possibly following up on
linux-kernel@vger.kernel.org.

Ok, thanks for that hint, I've run memtest and it gives me a lot of
error-counts!! Thats it!
Sorry, next time I'll get sure its not a hardware problem before posting
a message.. 

Thanks a lot!!!
sebastian